Introduction :
There's got to be a better way to test heatsink/fans than playing "Quake". Of course, when I began shopping for the best cooler for my new Tbird box, I really didn't know what I was in for. I especially wasn't ready for the widely differing views as to who made the best cooler...or actually if there is a "best" cooler at all.

Let me get you up to speed: I decided last spring to build myself what I considered to be my "ultimate" computer. My wish list included the fastest CPU I could get (1.2 GHz at the time), at least 2, 40 GB, 7200 RPM, ATA-100 drives with an ATA-100 RAID controller, 512 MB of PC2100 DDR memory, 400 Watt PSU, a 24" Trinitron wide aspect-ratio monitor, a dual-head video controller, and a case big enough for me to live in when my wife saw the bills start coming in.

In my quest to find the fastest processor and the appropriate motherboard and case I started checking out review sites and shopping sites and I started running across the word "overclocking" quite a bit. Although I remember my mother telling me quite sternly, albeit 40 years ago, not to hang around places like that, I nevertheless found the subject of "OCing" quite interesting. I particularly enjoyed the innumerable discussions about the two most common topics: cooling problems and the blue-screen-of-death. Since I am not a "gamer", the only reason I can come up with to overclock a perfectly good CPU is the same reason Sir Edmund Hillary climbed Mt. Everest. Doesn't make any sense to me either.

So, the most pressing problem I had was "how the heck am I going to cool this CPU that I'm going to kick right in the pants with a huge dose of voltage and front side bus"? Of course, the answer was to add the world's best heatsink/fan unit to my already prohibitively expensive wish list. And finding "the best" was going to be a piece of cake....or was it.

I managed to locate dozens, maybe hundreds of reviews and articles and advertisements all about heatsinks and fans, all shapes and sizes, some aluminum or silver or even gold. I read about fans that were so loud you'd need your fillings replaced if you sat in the same room. What I could not find was a simple ranking of all of the popular units on the market, with data in a format that I could use. I found that all of the current reviews published by the current crop of reviewers used pretty much the same procedure: get a bunch of HSF units, maybe 4 or 6 or 8 of them, and try them all out on the same computer. Play "Quake for an hour or two, or run Prime95 for a while, and check the temps. Well, that's not too good. All that does is tell me which of that particular batch of HSFs works best on his computer..not my computer...and certainly not your computer. And if that particular reviewer makes another run of tests with 6 more HSFs at a later date, the results don't mean a thing when compared to the prior batch. That's because he may not have the same computer. And, if he does, it's a few months older, and maybe not in the same prime condition it was in earlier. What complicates matters even further, is the fact that no two reviewers use the same methodology or terminology. There is no way any rational person can make an informed judgement about which heatsink/fan unit to buy.

So I decided to stick my 2 cents in and develop a method of testing these indispensible devices in a manner that will allow all of us to compare every HSF with every other one tested, even if the tests are a year or more apart. My idea was to design and fabricate a "test bench" that would dispense with the biggest variable: the computer. I wanted a device that would simply tell me how much heat energy a heatsink/fan could transfer. I figured it would be fairly simple. all I really wanted to do was obtain a measurement of either BTUs (British Thermal Units) or Calories. Everything else was totally irrelevant. If I can show myself, or you, that I have an accurate measure of just how much actual heat energy is moved away from the contact surface of the CPU, and I can obtain that measurement in a manner that is both simple and repeatable, then I can easily compare the results from 2 or 3 or 100 tests over whatever time period the testing covered.

Getting Started:

Fig. 1

Figure 1 is the business end of the unit. It is simply a bare Socket 462, manufactured by AMP/Tyco, and provided to me by the courteous sales staff at Sager Electronics. Embedded in the socket is a "Subminiature V-Tuned" waterblock purchased from the guys over at Overclock-watercool.com. Of course, I'm using it in a way that they never intended. The reason I chose this particular unit is its size: it isn't much bigger than a CPU die, so it will mimic actual conditions quite nicely. A small section of Lexan (3/32" thick) is being used to approximate the thickness of the CPU, with the block about 0.010" above the surface.

 

Fig. 2

Figure 2 is the thermometer/fan switch/power connection box. This is just a little Radio Shack project box that encloses all the wiring and switches (one for the fan and the other switches the two thermometer sensors). The thermometer is a nifty little unit that I purchased from Lascar Electronics. It is quite easy to mount and connect, and as far as I can tell it is fairly accurate.

 

Fig.3

Figure 3 shows a small, 12 volt "bilge" pump that I'm using to circulate 85C water through the waterblock/socket assembly. The pump is connected to the 13.8V power supply unit in the photo. Believe it or not, that power supply was the "AC adapter" for my first cell phone-20 years ago. The pump is simply an off-the-shelf unit purchased from W.W.Grainger.

 

Fig. 4

Figure 4 shows the whole rig ready for action. Note the K-Mart expanded polystyrene bait bucket. It might not be very pretty, but it gives me a watertight outer container with an R value of about 6. And I got all this beauty and functionality for $2! As you can see, all the water circulating apparatus is enclosed within the insulated container to minimize heat loss.

 

Methodology:

Before I start, let me address some issues that will become "big" issues to those manufacturers whose products rank lower than they think they should:

1. The waterblock, while very small, is still about 50% larger than an actual CPU die.

2. The apparatus has a certain amount of "heat loss" to the surrounding space, which is minimized by the insulation, and further reduced by an inner container that acts as a primitive "thermos".

3. The pump generates some heat when it is operating, which is unavaoidable.

I have made no attempt to quantify these three factors simply because they are constant and apply to every test and every tested device. The playing field is "level"...it just might be at a slightly different "altitude"

Our plan is to mount each HSF we get and run it for 30 minutes with an initial water temp of 185F (85C). At the end of the 30 minute time period, the temperature will be recorded and the BTUs and Calories computed. Of course, I'll probably monitor the temperature every 2 minutes or so, just so I can make a few fancy graphs, which mean basically nothing.. The bottom line is the final water temperature. This test is really designed to measure just one factor and produce one result: How fast can a particular HSF move heat energy from one place to the next. As an added bonus, we will report on such factors as noise, appearance, price, and mounting clip design, all of which will be evaluated in a purely subjective manner.

Finally, we will cheerfully test any HSF unit that we get. We'll test units supplied by manufacturers, we'll test units supplied by other reviewers (that way they can concentrate on the "subjective" stuff), and we'll test units that we buy (if we can't get them for nothing, of course). Our first 3 evaluations will appear here on 3DVelocity in the coming weeks. Get ready for a few surprises!